The AI Trade Has Entered Its Unleveraging Phase: Goldman's Structural Signal

The protocol does not lie; the interface does. In markets, the interface is the momentum factor, the hedged portfolio, the narrative-driven valuation. The protocol is the underlying flow of capital and the physical reality of infrastructure. On August 23rd, Goldman Sachs sent a clear signal through that interface. The AI trade is not dead. But the method by which market participants extract value from it has fundamentally changed. This is a critical juncture. We are witnessing a transition from the beta phase of the AI trade to a phase of alpha selection. The era where buying any AI-adjacent equity guaranteed outsized returns has ended. The new era demands a surgical understanding of which layers of the stack are actually generating revenue. The narrative of 'AI' as a monolithic sector is fracturing into its constituent parts: compute, memory, storage, software, and energy. For a sector that has been defined by its homogeneity in the eyes of capital, this is a profound shift. The leverage that fueled the initial rally is being unwound. The signal is clear for those who read the balance sheets and the risk models, not just the headlines. The era of passive allocation to the AI theme is over. The most significant data point from the Goldman Sachs analysis is the reversal in the momentum factor. Software has replaced semiconductors as the largest weight in the three-month momentum portfolio. This is not a trivial rotation. It is a declaration that the market's pricing power has moved from the hands of the infrastructure builders to the application layer. Let's call it what it is: the market is finally starting to price in the profit. The build-out of the infrastructure was phase one. The utilization of that infrastructure is phase two. Semiconductors, the quintessential 'picks and shovels' of the AI gold rush, have entered the short portfolio. This is the most direct signal of the de-rating. The market is not saying AI chips are obsolete. It is saying they are over-owned and over-priced relative to the near-term earnings trajectory. The challenges are mounting. The US export controls are limiting the addressable market. The large cloud service providers are scrutinizing their capital expenditures, and the rise of custom ASICs from the hyperscalers is chipping away at the monopoly of the incumbent. Goldman Sachs's identification of the storage and data center sectors as 'tactically most attractive' is a call to action. The logic is that the profit recovery is not yet reflected in the stock prices. This is a signal of a critical technical shift. The AI workload is moving from the training phase to the inference phase. Training is a concentrated, batch-driven operation. Inference is continuous, distributed, and power-hungry. This shift creates a different demand profile. To own the chain is to own the history. The history of AI's value transfer is now being written in storage. The massive vector databases, the model weights, and the KV cache systems needed to serve inference requests at scale require a storage capacity that dwarfs the initial training datasets. The value is moving from the computational state to the state layer. The data that is generated, stored, and retrieved in the inference process is the new gold. I have spent the last two weeks dissecting the storage market, and the numbers are clear. The HBM (High Bandwidth Memory) supply is sold out. The data center REITs are reporting rising occupancy and increasing power pricing power. This is not a cyclical uptick; it is a structural demand shift. The recovery is real, but the market is still pricing these companies as if they are stuck in the 2019 era of steady, but unspectacular, growth. The disconnect is an opportunity. However, the contrarian must look at the blind spots. The Goldman Sachs recommendation is a macro call. It is not a security checklist. The 'profit recovery' in storage is real, but it is not uniform. The market is treating 'memory' as a monolith. We have to make a clear distinction between the HBM, which is a scarce, high-margin product, and the NAND flash, which is a commodity with a history of price volatility. The recovery is driven by HBM, not the traditional SSD market. A blind investment in the entire storage index could lead to the same problem as the broad AI trade. The second blind spot is the speed of the rotation. The capital is moving to the European and Japanese banks, the gold miners, and the copper producers. This is a signal of the market looking for 'value' outside the tech complex. But copper, in particular, is an indirect bet on AI. Copper is the blood of the electrical grid. The AI data center power demand is not just a supply issue. It is a geopolitical and physical infrastructure issue. The capital moving into copper is a smart, indirect play on the AI revolution, but it is also a play on the global electrification story, which has a much longer duration. Let me be clear on the risks. The AI trade is still levered, albeit less so. The Goldman data shows the high-beta momentum portfolio fell 12% in a week and the AI hedge fund portfolio dropped 10% in five days. The deleveraging process is not complete. The next catalyst is the earnings report from the semiconductor giant in late August. The guidance, not the top-line revenue, will be the deciding factor. If the guidance is conservative, we will see another leg down. If it is strong, the rally could resume. The September industry conferences are the other key event. These are the moments where the roadmap is laid out. I am looking for one specific signal in the September conferences: the mention of 'inference at the edge'. If the conversation moves from the data center to the edge, the entire value chain changes. It would validate the shift from semiconductors to software. This is not a signal to exit the market. It is a signal to re-enter with a surgical precision. To own the chain is to own the history, but the history is now being written by the storage and the software. The market is finally rewarding the entities that are converting the AI promise into revenue. The AI is not ending. It is growing up. Certainty is a bug in a stochastic world. The only certainty is that the interface will continue to mislead us, and the protocol will continue to confirm the truth. The protocol of the AI economy is the revenue and the energy demand. The interface is the stock price. I will continue to read the protocol, not the interface. The takeaway is not a forecast of a crash. It is a forecast of a differentiation. The question for the investor is not whether to hold AI, but which layer of the stack to hold. The answer is moving from the input layer to the output layer. The market is not ending. The market is maturing. The infrastructure of the AI is now, and the infrastructure of the storage and data is the memory of the future. The silence before the block confirms the truth. The block is the revenue. The truth is the profit. And the market is finally learning to read it. We build in the dark to light the public square. The market has been in the dark. The Goldman signal is a flashlight. It illuminates the path from the compute to the memory. It is the investor's responsibility to walk the path with a clear understanding of the technical fundamentals, not the hype. The market will reward the careful, but the market will punish the careless. The time for precision is now.